How does the autonomous treasury fix

The advent of the sovereign treasury has ignited a competitive push, with aggressive industry targets. Not all companies will want to grow at the same pace.

The shift to an autonomous treasury is reshaping the world of corporate finance, driven by new strategies and technologies – from self-healing cash forecasting to AI-powered liquidity engines – that are replacing legacy systems and maximizing yield.

To fully realize the potential, corporate finance leaders are strategically investing in key areas that will accelerate the transition. Sayantan Chakraborty, head of digital payments at Fiserv, says the next phase of the sovereign treasury will be defined by three investment-focused areas. “Treasurers no longer lack visibility; they lack widgets that can act on that visibility in real time,” he says. “The difference is not analytical. It’s execution.”

Although agentic AI can forecast cash positions and draft funding instructions, notes Chakraborty, current corporate infrastructure often runs in batch mode. The first essential missing link is comprehensive, real-time cash positioning, second, it is combined with rules-based, just-in-time movement of funds across multiple payment rails – including instant and traditional – and third, the integration of new features like tokenized deposits and programmable payments.

However, technological travel still requires human expertise. And Chakraborty recommends building off legacy ERP systems rather than waiting for complete modernization.

“Think of it as an AI-powered autopilot added to an old cockpit,” he says. “Policies are implemented, actions are executed, and audit trails are preserved, all under the supervision of a trained cockpit and cabin crew, without a full-core replacement on day one.”

Chakraborty argues that the era of multi-year, large-scale upgrades is over. Instead, the best approach is to implement a lightweight, 24/7 automation layer to handle real-time balances, rules, and payments.

As instant payment rails and real-time reporting become more widespread, Chakraborty predicts that the current practice of pre-funding accounts before cut-off will become obsolete. Instead, “Agent AI will move the treasury from once-a-day instructions to continuous, just-in-time funding: as soon as execution matches intent at all levels.”

This change will impact the float, thereby reducing the idle-balance float and prompting banks to focus their earnings on 24/7 clearing services, intraday credit and real-time liquidity.

Siemens, a leader in autonomous treasury, adopted JPMorgan’s Programmable Payments feature (formerly Onyx, now Kinexis) in late 2023. Siemens moved into advanced programmable payments using the blockchain-based ledger, JPM Coin. It allows their bank accounts to manage cash autonomously and execute transactions based on pre-defined rules. Addressing the inefficiency of idle pre-funded balances, Siemens implemented a fair timing mechanism. Funds are transferred to a specific account only when payment is due. If the balance drops below a set threshold, the system autonomously withdraws funds from the central cash pool, enabling Siemens to operate with almost zero balance in local accounts.

“In my experience, the biggest challenge is not technology, but a change in mindset in finance and treasury,” says Heiko Nix, global head of cash management and payments at Siemens. “For almost every technical problem, there is a solution. But simplifying the underlying processes and changing the way people think about treasury and its role takes significantly more time and effort. In practice, you do not need to convince everyone at once, what matters is building enough momentum in the organization to enable real change.”

A ‘forward looking control tower’

John Stevens, senior vice president, global head of capital markets, financial institutions and working capital at Kiriba, argues that AI creates a strategic opportunity.

“AI can transform working capital management from a retrospective reporting function to a forward-looking control tower,” he says. “Instead of focusing on past events, you can adapt to the future in real time. That’s because tasks that previously required manual, analog effort, or required analysts to spend long periods of time consolidating reports, can now happen instantly. This real-time capability allows for significantly more intelligent and timely decisions.”

Companies still need to work closely with vendors to build AI safely, he warns: “We don’t see a single out-of-the-box ‘autonomous’ product replacing the diversity of fiscal needs.” He predicts that the future will be “composable”, although it is important to be precise about what this means.

While Kyriba App Studio serves as an extensibility layer for building custom integrations and workflows on the Kyriba platform, Stevens emphasizes that it is not an agent-building toolkit. The agentic AI layer is TAI, which provides Kiriba-developed agents with “a clear human in the loop posture.”

They argue that using third-party models does not automatically make an AI tool less intelligent and simply using in-house models does not automatically make it more intelligent.

“In treasury, the deciding factor is whether AI can be used safely and consistently in a regulated environment,” says Stevens. TAI is not in a position to avoid external LLM. “We use a leading external model [Anthropic’s Claude] Within a controlled, governed deployment. The difference is the wrapper around the model: strict limits on what data it can access, clear rules on what it is allowed to do, and a full audit trail of activity.

In practical terms, this means AI can help generate insights – summaries, clarifications, flag anomalies, scenario descriptions – while anything that could impact payments, liquidity or risk remains under platform control, approval and policy-driven workflows.

“So it’s not a binary choice between open and sovereign,” he said. “Some organizations will need sovereign options for policy or jurisdictional reasons, but most regulated treasuries are looking for governed AI: robust models that are secure, auditable and designed to have real operational control.”

Redefining Corporate Finance

The potential benefits to the treasury have ignited a competitive push for autonomy, with aggressive industry targets and a race for “fully autonomous” platforms.

HighRadius recently updated its Agentic AI platform with the goal of achieving more than 90% automation for the CFO office by 2027. The initiative involves deploying AI agents across accounts receivable, payable, treasury, close and consolidation across six product suites and 20 products. The release of 186 Agentic AI agents, announced last February, moves HiRadius closer to the “fully autonomous platform vision” it first announced in 2019, with cash applications and cash forecasting already demonstrating 90% touchless automation.

HighRadius prioritizes “measurable value creation”, which it validates with customers through Mutually Agreed Success Criteria (MASC). This value is delivered through automated agents, which aim for more than 90% automation, and assistive agents, designed to triple user effectiveness.

CEO Shashi Narahari sees agentic AI as an interim step toward HiRadius’ goal to ensure all of its products are “fully autonomous” by 2027 – defined as a more than 90% touchless end-to-end process. Narahari emphasizes the important nature of this goal, to the extent that failing to achieve it would spell the end of the company.

What about mid-tier banks that may not want to pursue radical change? For them, Chakraborty advises that a single, reliable orchestration endpoint is better than multiple disparate APIs.

“What it requires is a real-time balancing plus payment execution API,” he says, “exposing positions, limits and instant movement through a single, flexible interface. This is what lets AI-powered treasury systems act not just as analysts, but as agents.” He further added that where possible, it is also beneficial to integrate such a process with the token deposit movement.

That said, the journey toward autonomous treasury, led by leading companies like Siemens and driven by the rapid development of agentic AI, is fundamentally redefining corporate finance.

This shift is not just about incremental efficiency gains, but is being seen as a strategic imperative to maximize yield, secure real-time liquidity and go beyond the constraints of legacy systems. Corporate treasurers who are embracing change are attracted by the promised strategic roadmap to a future-proofed role. For the financial institutions that serve them, Autonomous Treasury is an urgent call to align their offerings with a new era of consistent, intelligent and timely financial regulation.